Koustuv Sinha
Papers
1
Total Citations
3
H-Index
1
About
Koustuv Sinha is an AI researcher working at the intersection of self-supervised learning, video understanding, and world modeling. His most notable recent contribution is **V-JEPA 2** (2025), a landmark paper exploring how AI systems can learn to understand, predict, and plan by observing the world — a fundamental challenge in modern artificial intelligence. This work advances the Joint Embedding Predictive Architecture (JEPA) framework by combining internet-scale video data with robot trajectory data to build models capable of both perception and action-oriented reasoning, representing a significant step toward general-purpose AI that learns largely through observation rather than explicit supervision. Though V-JEPA 2 is early in its citation trajectory with 3 citations at time of writing, its scope and ambition place it among the more consequential contributions in self-supervised video modeling — an area rapidly gaining centrality in the AI research community. Sinha's work sits at a compelling frontier where representation learning meets embodied intelligence, making his research particularly relevant to students and practitioners interested in scalable, data-efficient approaches to building AI systems that genuinely understand the dynamic visual world.
Research Focus
Key Achievements
Top Papers
- 1